Google Ads metric changes in 2026: view-through conversions, the 10-second engagement threshold and conversion windows

Your engagement rate halved in October. Revenue didn’t move. Or view-through conversions jumped by a third while your CRM recorded exactly the same number of deals as last month. Neither is an auction anomaly, and neither is “Google throttling you.” In most cases these are Google Ads metric changes at the level of definition: the number moved because the counting rule changed, not because people behaved differently.

Through 2026 Google rewrote the definition of several reported metrics — usually with no blog post, no press release, just a quiet edit to a Help page or a line in release notes. This article gives you a working procedure: how to tell a redefinition from a genuine performance drop in about fifteen minutes, how to keep a definitions log so these breaks stop eating your reporting, and what to do with dashboards, KPIs and bidding once a data series is officially no longer comparable with itself.

Google Ads metric changes: what actually got redefined in 2026

These are the changes already sitting in most accounts’ reports. For each one, note whether there is an official effective date — that determines how much weight it can carry in a post-mortem.

1. The engagement threshold moved from 5 seconds to 10

An archived copy of Google’s engagement reporting Help page from 15 March 2026 set the video threshold at 5 seconds watched. The current page requires 10 seconds, the full ad if it is shorter, or a click. Formats that had no separate entry in the March copy — Shorts and masthead — now have their own rules. Bumper and non-skippable in-stream still count clicks only.

Format	March 2026 wording	Current wording
Skippable in-stream	5 seconds watched	10 seconds or the full ad
Video app promotion	10 seconds	10 seconds (unchanged)
In-feed video	5-second rule	separate entry, 10 seconds
Shorts, masthead	no separate entry	own rules added
Bumper, non-skippable	clicks only	clicks only

Important: Google published neither an effective date nor an impact estimate. The arithmetic, however, is not in dispute: the numerator (engagements) shrinks while impressions stay the same, so engagement rate falls and cost per engagement rises at identical spend. Any “30–50% drop” figure circulating in industry roundups is a directional benchmark, not a Google statement — measure the actual size of the step in your own account.

2. GA4 dropped the fixed 3-day engaged-view window

On 11 August 2026, a line in the Google Analytics release notes — no blog post, no announcement — replaced the hard-coded three-day engaged-view lookback with any whole number from 1 to 30 days. Click-through windows moved from six presets (1, 7, 14, 30, 60, 90) to any whole number from 1 to 90. The setting lives under Advertising → Conversion management → more options → Settings.

The consequences run deeper than the setting suggests. Widening the engaged-view window credits YouTube views with purchases made weeks later, and it is genuinely unclear whether those conversions are incremental or simply reassigned from search and direct. Earlier documentation indicated lookback changes apply from the edit date rather than retroactively, which means your series breaks and any year-over-year comparison silently mixes two definitions.

3. Demand Gen view-through conversions

Industry roundups report that from October 2026, Demand Gen view-through conversions count rendered impressions instead of Active View. The practical effect is that October VTC numbers rise with no underlying improvement. There is no official announcement as of publication, so treat it as a hypothesis to confirm in your own data: if VTC stepped up on a single day while clicks, click-through conversions and revenue stayed flat, you are almost certainly looking at a recounted base. The billing and attribution mechanics of this campaign type are covered in our breakdown of Demand Gen CPM billing and view-through conversions.

4. Adjacent changes worth logging

In September 2026 Smart Bidding calibration guidance shifted toward “50 conversions or three learning cycles,” and the Data Strength Uplift metric entered reporting. Neither breaks a historical series directly, but both change the thresholds you make decisions against — so both belong in the same log.

The rule that saves hours of investigation: any step change in a chart with no matching step change in money is a redefinition candidate, not an incident.

Why this hurts reporting more than it hurts campaigns

A redefinition almost never touches the auction. Users behave the same, bids are the same, the landing page is the same. What breaks is the measurement layer — and then everything built on top of it:

• KPIs and bonuses. If the team target is “engagement rate above X,” a 10-second threshold wipes out attainment even though nobody did anything worse.
• Year-over-year comparisons. October 2026 against October 2025 is two different metrics wearing the same label.
• Automated rules and alerts. A rule that pauses campaigns on a 25% engagement drop will fire across the account and switch off healthy campaigns. Safe patterns are covered in our guide to automated rules in Google Ads.
• Bidding. A changed conversion window changes the data the algorithm learns from — see conversion windows and conversion lag.
• Client trust. The expensive one. Explaining a drop after the question is always worse than annotating the chart the day before it.

The fifteen-minute test: definition or result

Run this sequence whenever an unexplained step appears. Stop as soon as three signals point the same way.

Step 1. Shape: step or slope

Chart the metric daily over 60–90 days. A redefinition almost always produces a step: one level before date X, another after, with no transition. Genuine deterioration — competitors, seasonality, a broken tag, a slow landing page — usually produces a slope or a jagged line with partial recoveries.

Step 2. Blast radius

Check whether every campaign of that type is affected, and if you have access, every account under your manager account. A definition change hits everyone using the format simultaneously; a campaign problem hits selectively. Cross-account comparison through your manager account (MCC) is the fastest way to answer this.

Step 3. Neighbouring metrics

Real drops travel in packs. If engagement fell because the creative got worse, CTR, average watch time and conversions usually moved too. A redefinition touches exactly one metric and its derivatives (cost per engagement) while impressions, clicks and conversions stay put. This is the single strongest discriminator.

Step 4. An independent source

Check data Google did not redefine: raw events in your GA4 BigQuery export, orders in the CRM, server logs. If those are flat, the question is settled. The structural reasons platforms and analytics disagree are covered in Google Ads vs GA4 conversion discrepancies.

Step 5. Your own change history

Rule yourself out before blaming the platform: bid edits, a changed conversion action, a new landing page, an edited attribution window. All of it is visible in change history — the workflow is in using change history to diagnose performance drops.

Step 6. Match the date to the release timeline

Line the step date up against platform change logs. A match within a day or two, combined with signals from steps 1–4, is your answer. The absence of an official announcement does not weaken the conclusion — several 2026 changes arrived as silent Help-page edits.

Signal	Metric redefinition	Real performance change
Chart shape	Sharp step on one date	Slope, waves, partial recovery
Scope	Every campaign of that format	Specific campaigns or ad groups
Adjacent metrics	Unchanged	Move together
Money	Revenue and orders flat	Revenue follows
External source	Confirms nothing happened	Confirms the drop
Change history	Empty on that date	Usually shows an edit

A worked example with real numbers

A hypothetical account running YouTube video campaigns at a steady ~$9,000 a week. The buyer opens the report and sees what looks like a disaster:

Metric	Week before	Week after	Change
Spend	$9,040	$8,980	−0.7%
Impressions	3.10M	3.08M	−0.6%
Engagements	214,000	121,000	−43%
Engagement rate	6.9%	3.9%	−3 pp
Cost per engagement	$0.042	$0.074	+76%
Click-through conversions	412	408	−1%
Revenue (CRM)	$61,300	$60,800	−0.8%

Run the test. Shape: a step on one specific Tuesday, not a slope. Scope: every video campaign, including two nobody has touched in six months. Adjacent metrics: impressions, clicks, conversions and revenue all flat; only engagements and their derivative moved. Change history on that date: empty. Conclusion: definition, not result. Response: annotate the chart, send the client the single table above, rebase the engagement rate target over the next two weeks — and change nothing in the campaigns.

Contrast that with a genuine collapse: spend flat but impressions falling or getting more expensive, CTR sagging alongside engagement, conversions following within a few days, and a bid edit or a new ad group sitting in change history two days earlier. That one needs fixing, quickly.

How the response differs by campaign type

• Video and Demand Gen. The hardest hit, because engagements and view-through conversions live here and creative decisions are often made on them. This is where “fixing something that isn’t broken” is most likely.
• Search. Barely touched directly, but exposed indirectly: if your cross-channel budget model weights channels by engagement, the distortion travels.
• Performance Max. The risk is blending — reporting is aggregated, so a step in one channel can read as a whole-campaign decline.
• Campaigns with offline conversions. The most sensitive to window changes, because a window edit also changes which uploads are counted. Always log the edit date.

A definitions log that takes five minutes a week

The cheapest insurance against repeat investigations is a table owned by whoever owns reporting. Five columns, nothing more:

1. Effective date — or the date you observed the step, if no official date exists.
2. What changed — one line: “engagement threshold 5 → 10 seconds.”
3. Metrics and formats affected — so nobody has to guess later which charts are not comparable.
4. Retroactive? — whether history was recounted. GA4 window changes, for instance, apply from the edit date, so history keeps the old definition.
5. Source — a link to the Help page, the release note, or your own screenshot of the chart.

Two rules that go with the log. First, never change a conversion window and a bidding setting on the same day — you will not be able to separate the two effects afterwards. Second, log your own changes too: attribution model swaps, new offline uploads, a move to server-side tagging. For a mature account those break the series exactly as hard as a platform edit does.

Rebuilding dashboards so the break is visible

A dashboard’s job is not to hide the step. It is to make it obvious.

• Vertical annotations on redefinition dates. In Looker Studio this is a small date/label table used as a reference line overlay. Base reporting assembly is covered in Looker Studio reporting for Google Ads.
• Break the line instead of drawing through it. Two series — “old definition” and “new definition” — are more honest than one continuous line that invites an invalid comparison.
• Keep anchor metrics beside it. Always place a stable anchor next to a redefined metric: impressions, clicks, revenue. That answers half the questions before anyone asks them.
• Custom columns, not spreadsheet patches. If you need a reconstructed “old” version of a metric for comparability, build it as a column so the logic survives staff turnover — see custom columns and the report editor.

What to do with targets and bidding

Rebase the KPI. If a target was tied to engagement rate or VTC, establish a new baseline from two to four clean weeks after the change and manage against that. The old number no longer means what it meant.

Audit which conversions feed bidding. A changed window changes what the algorithm sees. If you widen a window, expect reported conversions to rise and do not read that as improvement until incrementality confirms it.

Let the strategy settle. After any change that affects conversion flow, a sensible benchmark is to leave settings alone until you have 50 conversions or three learning cycles. Decisions made inside that window are usually explained by noise.

If you need truth rather than a report, measure lift. Redefinitions do not distort a lift test, because both groups are measured under the same definition — see Conversion Lift for Search and Performance Max.

Five mistakes that cost the most

1. Fixing a campaign that isn’t broken. Engagement “drops,” the buyer swaps creatives and targeting, and now something that worked is broken and learning has reset.
2. Reporting year-over-year without a caveat. To a client that reads as the team regressing.
3. Retroactively “repairing” history. Only possible if you know the exact formula. You usually don’t — show the break instead.
4. Changing several things at once. Window plus attribution model plus a new conversion action makes every subsequent movement unattributable.
5. Not writing it down. In three months nobody remembers why October has a step, and the investigation starts over.

This week’s checklist

• Export 90 days of daily data: engagements, engagement rate, cost per engagement, VTC, conversions, revenue — and mark every step.
• Run the six-step test on each step you find.
• Start a definitions log and date-stamp every one of the Google Ads metric changes you find — starting with the engagement threshold and the GA4 window.
• Audit automated rules and alerts for thresholds tied to redefined metrics.
• Add dashboard annotations and brief the client before the monthly report goes out.
• Set new KPI baselines after two to four weeks of clean data.

If you are building a measurement process from scratch, the fundamentals are collected in the PPC Rebels Google Ads guide, and more account-level breakdowns are published in the PPC Rebels blog.

FAQ

How do I know quickly whether a metric was redefined rather than dropped?

Three signals at once: a sharp step on a single date, every campaign of that format affected, and impressions, clicks, conversions and revenue unchanged. That combination is almost always a counting-rule change.

Does Google have to announce these changes?

No. Several 2026 changes arrived with no blog post or press release — only a Help-page edit or a release-note line. Your own log is more reliable than waiting for an announcement.

Is the engagement threshold now 10 seconds everywhere?

For skippable in-stream it is 10 seconds, the full ad if shorter, or a click. In-feed, Shorts and masthead now have their own Help-page entries; bumper and non-skippable still count clicks only. Check the current page for the specific format before planning video.

Did the threshold change affect billing or bidding?

The Help page does not say. The safe position is to treat it as a reporting change and verify actual spend and cost per conversion in your own before/after data rather than drawing conclusions from the engagement metric alone.

Should I change my GA4 conversion window now that 1–30 days is available?

Only if you have a hypothesis you are testing and accept the series break. Changes apply from the edit date, history is not recounted, and comparability with other accounts and periods is lost.

Why did view-through conversions rise when I changed nothing?

The likely cause is the October 2026 change to how Demand Gen counts them. Verify it: if VTC stepped up while click-through conversions and revenue stayed flat, it is a recount, not a result.

Should I restate history under the new definition?

Usually not. The exact restatement formula is rarely known, so you end up with a tidy but false series. A broken line plus an annotation is more honest.

How do I explain this to a client without sounding defensive?

One slide: the date, what changed, which metrics are affected, proof that money is unaffected (revenue, CRM orders), and the new baseline. Delivered before the question, it reads as control rather than excuse.

Can a redefinition affect Smart Bidding learning?

Only if conversion flow changed — the window, the set of conversion actions, or the data source. Purely reported metrics like engagement do not affect learning, though they often trigger harmful manual edits.

How often should I check platform change logs?

Weekly, plus a mandatory check before the monthly report. That cadence is enough for the annotation to land before the client’s question does.

Which metrics make good anchors on a dashboard?

The ones whose definitions change least and that the business understands: spend, clicks, orders and CRM revenue. Next to those, a step in any derived metric is immediately readable.